Designing an adaptive learner model for a mathematics game

The RAIDING Project (Researching Adaptivity for Individual Differences in Number Games) aims to develop a game for 7-8 year olds, to develop their times tables and number bond skills. One of the design principles of the project is to implement a level of adaptivity into the game, so that the difficulty of the mathematical content adapts to the player's current level of arithmetic fluency. A learner model has been developed to enable the game to use previous gameplay performance to calculate the player's current level of arithmetic fluency, and thereby provide new tasks at an appropriate level of difficulty. A second design principle is to decouple the mathematical difficulty from gameplay rewards, so that progress in the game is achieved through time and effort rather that solely as a result of mathematical achievement. We predict that these two design principles will produce games that are motivating and help players to experience flow. This paper describes and discusses our adaptive implementation, and our approach to decoupling of mathematical learning from rewards. We evaluate the success of the game to date and consider scope for potential development and improvement. We also show how the analytical data produced by the learner model has been used to identify unhelpful in-game behaviours and adapt the game design. A future goal of the project is to explore whether the adaptivity of the learner model can be expanded to include gameplay ability (including elements hand-eye coordination and response times) and allow for separate dynamic adjustment of (non-mathematical) difficulty. We are particularly interested in investigating the affordances of such a "two-axis" flow in the game.